Yeah, but what has ML ever done for neuroscience?

This question has been going round the neurotwitters over the past day or so.

Let’s limit ourselves to ideas that came from machine learning that have had an influence on neural implementation in the brain. Physics doesn’t count!

  • Reinforcement learning is always my go-to though we have to remember the initial connection from neuroscience! In Sutton and Barto 1990, they explicitly note that “The TD model was originally developed as a neuron like unit for use in adaptive networks”. There is also the obvious connection the the Rescorla-Wagner model of Pavlovian conditioning. But the work to show dopamine as prediction error is too strong to ignore.
  • ICA is another great example. Tony Bell was specifically thinking about how neurons represent the world when he developed the Infomax-based ICA algorithm (according to a story from Terry Sejnowski). This obviously is the canonical example of V1 receptive field construction
    • Conversely, I personally would not count sparse coding. Although developed as another way of thinking about V1 receptive fields, it was not – to my knowledge – an outgrowth of an idea from ML.
  • Something about Deep Learning for hierarchical sensory representations, though I am not yet clear on what the principal is that we have learned. Progressive decorrelation through hierarchical representations has long been the canonical view of sensory and systems neuroscience. Just see the preceding paragraph! But can we say something has flowed back from ML/DL? From Yemins and DiCarlo (and others), can we say that maximizing the output layer is sufficient to get similar decorrelation as the nervous system?

And yet… what else? Bayes goes back to Helmholtz, in a way, and at least precedes “machine learning” as a field. Are there examples of the brain implementing…. an HMM? t-SNE? SVMs? Discriminant analysis (okay, maybe this is another example)?

My money is on ideas from Deep Learning filtering back into neuroscience – dropout and LSTMs and so on – but I am not convinced they have made a major impact yet.

“Firing,” by d. m. kingsford


I was roaming the streets of Denver during an ultra-long layover on Friday and ran into someone offering to write poems on the spot, on any topic. The topic: brains, neurons.

Merry Christmas, neuroscience community:

Firing by d.m. kingsford

Like a V-10,000,000
this thing, ordinary enough,
comprised of the same stuff
as everyone else’s,
making up a man of average intelligence,
kind, occasionally
(but his in-laws think he’s a
and basically fulfilled.

this thing is firing
on all cylinders, heat beat,
renal systems in check,
temperature ok, and
at this moment,
the frontal lobe bearing down on
a crossword puzzle.

The same as Stephen Hawking.

(Apologies for the loss of formatting.)


Studying the brain at the mesoscale

It i snot entirely clear that we are going about studying the brain in the right way. Zachary Mainen, Michael Häusser and Alexandre Pouget have an alternative to our current focus on (relatively) small groups of researchers focusing on their own idiosyncratic questions:

We propose an alternative strategy: grass-roots collaborations involving researchers who may be distributed around the globe, but who are already working on the same problems. Such self-motivated groups could start small and expand gradually over time. But they would essentially be built from the ground up, with those involved encouraged to follow their own shared interests rather than responding to the strictures of funding sources or external directives…

Some sceptics point to the teething problems of existing brain initiatives as evidence that neuroscience lacks well-defined objectives, unlike high-energy physics, mathematics, astronomy or genetics. In our view, brain science, especially systems neuroscience (which tries to link the activity of sets of neurons to behaviour) does not want for bold, concrete goals. Yet large-scale initiatives have tended to set objectives that are too vague and not realistic, even on a ten-year timescale.

Here are the concrete steps they suggest in order to from a successful ‘mesoscale’ project:

  1. Focus on a single brain function.
  2. Combine experimentalists and theorists.
  3. Standardize tools and methods.
  4. Share data.
  5. Assign credit in new ways.

Obviously, I am comfortable living on the internet a little more than the average person. But with the tools that are starting to proliferate for collaborations – Slack, github, and Skype being the most frequently used right now – there is really very little reason for collaborations to extend beyond neighboring labs.

The real difficulties are two-fold. First, you must actually meet your collaborators at some point! Generating new ideas for a collaboration rarely happens without the kind of spontaneous discussions that arise when physically meeting people. When communities are physically spread out or do not meet in a single location, this can happen less than you would want. If nothing else, this proposal seems like a call for attending more conferences!

Second is the ad-hoc way data is collected. Calls for standardized datasets have been around about as long as there has been science to collaborate on and it does not seem like the problem is being solved any time soon. And even when datasets have been standardized, the questions that they had been used for may be too specific to be of much utility to even closely-related researchers. This is why I left the realm of pure theory and became an experimentalist as well. Theorists are rarely able to convince experimentalists to take the time out of their experiments to test some wild new theory.

But these mesoscale projects really are the future. They are a way for scientists to be more than the sum of their parts, and to be part of an exciting community that is larger than one or two labs! Perhaps a solid step in this direction would be to utilize the tools that are available to initiate conversations within the community. Twitter does this a little, but where are the foraging Slack chats? Or amygdala, PFC, or evidence-accumulation communities?

5 things I learned on Sunday at #sfn16

Just some irrelevant facts.

1. Toadfish can sing by vibrating their swim bladder, and can vocalize independent of breathing.

2. Scientists working in Drosophila are starting to see reafferent signals (signals representing motor commands such as walking or singing) in tons of places, from sensory neurons on downward. This is one of the areas where Drosophila neuroscientists are way ahead of mammalian neuroscientists.

3. The computations that visual neurons perform can fundamentally alter in conditions that seem like they should be similar, such as light levels (this is not just adaptation to statistics).

4. There are conditions in which retinal neurons can be synergistic and not redundant. This is a bit of a controversy in the field, with the common consensus that ganglion cells have ~10% redundancy. Apparently this is not always the case!

5. Drosophila (fruit flies) have a spatial short term memory that has been located in the central complex (actually, I am not clear exactly where the anatomical structure is located: it may be just outside of the central complex.)

#sfn16 starts tomorrow in San Diego

I know many of you will be there! I am giving a talk on Sunday morning at 8am. Since everyone will be fresh-faced and excited for the conference, I am sure it will not be a problem to be there that early, right? Right? The talk will be in room SDCC 30B on a new method to for unsupervised analysis of behavior which I am really, really excited about (we have gone beyond what the abstract says so ignore that).

Papers for the week, 10/30 edition

Recurrent Switching Linear Dynamical Systems. Scott W. Linderman, Andrew C. Miller, Ryan P. Adams, David M. Blei, Liam Paninski, Matthew J. Johnson. 2016.

The Serotonergic System Tracks the Outcomes of Actions to Mediate Short-Term Motor Learning. Takashi Kawashima, Maarten F. Zwart, Chao-Tsung Yang, Brett D. Mensh, Misha B. Ahrens. 2016.

A visual circuit uses complementary mechanisms to support transient and sustained pupil constriction. William Thomas Keenan, Alan C Rupp, Rachel A Ross, Preethi Somasundaram, Suja Hiriyanna, Zhijian Wu, Tudor C Badea, Phyllis R Robinson, Bradford B Lowell, Samer S Hattar. 2016.

Numerical analysis of homogeneous and inhomogeneous intermittent search strategies. Karsten Schwarz, Yannick Schröder, and Heiko Rieger. 2016.

Unexpected arousal modulates the influence of sensory noise on confidence. Micah Allen, Darya Frank, D Samuel Schwarzkopf, Francesca Fardo, Joel S Winston, Tobias U Hauser, Geraint Rees. 2016.

Vision Drives Accurate Approach Behavior during Prey Capture in Laboratory Mice.
Jennifer L. Hoy, Iryna Yavorska, Michael Wehr, Cristopher M. Niell. 2016.

Amygdala and Ventral Striatum Make Distinct Contributions to Reinforcement Learning. Vincent D. Costa,Olga Dal Monte, Daniel R. Lucas, Elisabeth A. Murray, Bruno B. Averbeck. 2016.

Potent optogenetic inhibition of behavior with anion channelrhodopsins.  Farhan Mohammad, James Stewart, Stanislav Ott, Katarina Chlebikova, Jia Yi Chua, Tong-Wey Koh, Joses Ho, Adam Claridge-Chang. 2016.

Interstellate Magazine: The Art of Neuroscience


Caitlin Vander Weele has been curating a collection of stunning neuroscience images that really bring the brain to life (so to speak). Volume 1 is here and it is a beauty! You can also download the hi-res PDF if you’ve got a spare 210 MB lying around.

Some samples:



Posted in Art

Coherent dots as art


One of the more influential experimental paradigms in sensory neuroscience is the coherent random dots task in which small dots flicker in and out of existence, with some small number of them moving either left or right, like flecks of snow on a windy winter day. An animal – a monkey, a mouse, a human – is forced to say in which direction these dots are moving, a task which gets harder as the number moving in a coherent direction gets smaller. You can see an example here (which is uploaded in quicktime for some reason). Versions of this task have been adapted to other sensations like audition.

I was in Seoul recently and visited the Seoul Museum of Art. Filled to the brim with amazing installations, one caught my eye. Much to the chagrin of my non-neuroscientist companions, I became entranced by a vivid representation of these seminal psychophysics studies. Norimichi Hirakawa, consciously or not, has manifested this ‘random noise’ into a form that is somehow accessible to a broad audience. Think of the possibilities inherent in that the next time that you run an experiment.

(I took some videos but could not find a good way to embed them as youtube and vimeo both attempt to lossily compress it – which is difficult when you are literally compressing noise.)

RIP Roger Tsien, 1952-2016


Sad news today – Roger Tsien passed away one week ago.

Can anyone imagine biology today without GFP? And though he is best known for that – he did share the Nobel prize for GFP, after all – Roger Tsien did much more. One of my favorite stories about Roger (and he was a character; you couldn’t be at UCSD without one or two Roger stories) came from a talk he gave while I was there. He was asked to step in at a somewhat late moment to give a series of three one-hour lectures on his life’s work. It was one of the best talks I have attended, clear and insightful and packed with interesting anecdotes.

He begun describing his struggles as a graduate student, how he had a hard time doing electrophysiology. His solution? He created BAPTA, the calcium chelator that is the basis for neural imaging. Yeah: electrophysiology too hard? Just create the basis for calcium imaging. Much easier. (And he was quite honest; it was much easier for him.)

You should of course read that original BAPTA paper (this is under-appreciated!). Then of course there is his GFP paper, establishing a point mutation with much improved characteristics. Here is his Nobel lecture. I don’t think it’s an understatement to say that these are just the tip of the iceberg.

Monday Open Question: Does anyone actually know what a ‘reflex’ is?

It has been a while since I have done one of these…

As I was working on a fellowship application last week, I realized that I did not know whether what I was studying would count as a ‘reflex’ or not. What was the definition?  Is something not a reflex simply when we have a hard time mapping the input to the output? The canonical reflex arc kind of gives us one definition, but a good definition will be general – applicable to both mammals and non-mammalian creatures (dragonflies, antlions, worms). I asked on twitter and got some unsatisfying answers.

Is everything that is fewer than n synapses a reflex? What if those connections are mediated by state in some way (say, a peptide) so that some mapping from sensation to action depends on hunger, on mood, on whatever else? Are the only actions that are not reflexes those that are not dictated by sensory input…somehow? Is this just a lazy way to beat up on invertebrate neuroscientists?